The conventional paradigm of translation initiation in eukaryotes states that the cap-binding protein complex eIF4F (consisting of eIF4E, eIF4G and eIF4A) plays a central role in the recruitment of capped mRNAs to ribosomes. However, a growing body of evidence indicates that this paradigm should be revised. This review summarizes the data which have been mostly accumulated in a post-genomic era owing to revolutionary techniques of transcriptome-wide analysis. Unexpectedly, these techniques have uncovered remarkable diversity in the recruitment of cellular mRNAs to eukaryotic ribosomes. These data enable a preliminary classification of mRNAs into several groups based on their requirement for particular components of eIF4F. They challenge the widely accepted concept which relates eIF4E-dependence to the extent of secondary structure in the 5' untranslated regions of mRNAs. Moreover, some mRNA species presumably recruit ribosomes to their 5' ends without the involvement of either the 5' m(7)G-cap or eIF4F but instead utilize eIF4G or eIF4G-like auxiliary factors. The long-standing concept of internal ribosome entry site (IRES)-elements in cellular mRNAs is also discussed.